Accounting for territory and representative heterogeneity in sales force performance evaluation

A major challenge in evaluating, ranking and training sales representatives arises from the heterogeneity in both their territories and characteristics of the representatives themselves. For example, market potential and geographical coverage may vary greatly across sales territories, and yet many managers typically use only sales or profit outcomes without accounting for territory difficulty differences. This research describes the first use of a hierarchical Bayesian model for evaluating sales representatives that can directly account for all levels of heterogeneity and variation. In addition, we demonstrate how this framework can be used for ranking sales representatives, a ubiquitous practice for purposes of awards, bonuses, and promotions and for “gap analyses”, that is assessing areas in which the sales representative is not as adept at converting potential into sales.; We model sales as a function of an overall representative “effect” (a random intercept model), a product-specific representative “effect” and territory level factors. Simultaneously, we model the overall representative effect as a function of representative characteristics such as years in the territory. Consistent with research in the fairness and compensation literature, our models for ranking focus solely on factors that are beyond the control of the representative. Our computational approach, using Markov chain Monte Carlo techniques, generates posterior samples which allows us to answer a variety of managerially relevant questions; such as who are our top and bottom performers and how certain are we about their ranks? Other outputs of the model include opportunities for improving incentive compensation and allocation of training resources.; We compare results to standard approaches, for a data set obtained from a large producer of pharmaceutical products, and find improved out-of-sample fit as well as divergent ranking results.